Fault Injection with a New Flavor: Memetic Algorithms Make a Difference

摘要

During recent years we observe an arms race between new creative methods for inserting effective faults and designing new coun-termeasures against such threats. Yet, even analyses of an unprotected smart card pose a problem for an analyst assuming constraints in time (or consequently, in a feasible number of measurements). In this paper we present a new kind of algorithm capable of finding faults in the black box test scenario - memetic algorithm. This algorithm combines the strengths of the following three algorithms: genetic algorithm, tabu search and local search. Furthermore, the same algorithm can be used if the goal is simply a rapid characterization of the search space. We compare our algorithm with random search and exhaustive search approaches. Experimental results show that our memetic algorithm is substantially more successful in both, locating faults and characterizing search space, than the other known methods. In reaching both goals, our memetic algorithm uses less than 300 measurements.